The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill documentation references shell scripts for installing development tools from the vendor's domain and well-known providers, such as the infsh CLI (https://cli.inference.sh), the uv package manager (https://astral.sh/uv/install.sh), and Node.js managers from Vercel and other established sources.
[REMOTE_CODE_EXECUTION]: Installation commands utilize piped shell execution (e.g., curl ... | sh), a pattern used to streamline the setup of development environments using scripts hosted by the platform vendor and recognized technology services.
[COMMAND_EXECUTION]: The skill provides instructions for using the infsh command-line tool to perform various tasks like application scaffolding, local testing, and cloud deployment, which involve executing shell commands on the local system.
[PROMPT_INJECTION]: The skill provides a framework for applications that ingest and process external data, establishing a surface for indirect prompt injection.
Ingestion points: Applications receive untrusted user data via standard input models.
Boundary markers: The provided templates do not implement delimiters to distinguish user input from system instructions.
Capability inventory: Applications have capabilities including network access, file system interaction, and GPU compute.
Sanitization: Input validation is focused on schema types rather than content sanitization.

building-inferencesh-apps